> Alternative (slightly more compact) implementation of tv:
>
> tv=: (0 = {.) *./@, (0 < {:), 1 >: +/\inv
Nice way to factor out the (*./) redundancy. Thanks.
> Meanwhile, this is probably a bit long (almost 100 characters) to be
> called a one line implementation, but it seems to be in the right
> ballpark:
>
> ;"1(' - ',&.>' o+|'){~(+3*_1 1&E.)"1&.|:(*
> >./\.@:*)0,~2:`(}.@((i:,.]-1:) ~.))`((1,~]#_1:)"0)}~d
> -o
> |-o
> | |-o
> | +-o
> |-o
> +-o
> |-o
> +-o
> +-o
Argh. Reading this made me smile. I really like several of the ideas in here,
especially the use of boxing to select character pairs at the end. Also, the
(_1 1&E.) is such a compact edge detector, but unfortunately in this case, it's
not quite enough:
tp=: ;"1(' - ',&.>' o+|'){~(+3*_1
1&E.)"1&.|:(*>./\.@:*)0,~2:`(}.@((i:,.]-1:) ~.))`((1,~]#_1:)"0)}~
d=: 0 1 2 3 1
tp d
-o
|-o
| +-o
| + +-o
+-o
The problem is that detection of branch endpoints isn't locally determined; we
have to look arbitrarily far ahead. This is the "catenary problem" I mentioned.
I'm going to have some fun stealing your ideas.
Cheers!
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
